The world is experiencing an absolute frenzy observing the potential scope of the use of artificial intelligence systems through chatbots such as ChatGPT. But the reality is that this platform was only the beginning, here is ChatLLaMA from Meta, which promises to be a much more dangerous beast.
If they have tried the use of this type of platform, they will have verified that at first impression they are tremendously efficient and capable of generating quite robust pieces of information and that they could well pass for something that a human being articulated.
From chores, to diets, to grocery shopping lists, to imitating celebrities’ thinking as if having a chat with them, to writing entire essays, and even malicious code on the Dark Web, ChatGPT has proven that it can do quite a few things, not perfect. but way above average.
But with the arrival of ChatLLaMA the feeling is one: we are about to enter a new stage where different chatbot platforms powered by Artificial Intelligence will begin to circulate, each one more precise and powerful than the previous one.
This is ChatLLaMa: the new chatbot that would dethrone ChatGPT
It happens that, according to reports, Mark Tech PostNebuly, an Artificial Intelligence systems optimization company has just introduce ChatLLaMA, the first open source implementation of the LLaMA language model originally created by Meta, the company that brings together Facebook, Instagram and other platforms you probably use.
In essence, Mark Zuckerberg and his Fundamental AI Research (FAIR) team developed LLaMA with the aim of helping researchers, engineers and scientists to implement the use of AI to facilitate their work, but now with ChatLLaMA we see that it seeks to implement at purest ChatGPT style for any mortal to use.
At the moment this system is only available in Githubbut its potential would be very disturbing thanks to training parameters much higher than those of GPT-3 (where ChatGPT comes from).
And now they are open to everyone to grow even more thanks to this implementation based on reinforced learning with human feedback (RLHF).
By opening up its interaction with users, the model would improve substantially since ChatLLaMA is faster and easier to train thanks to its base architecture of 13 billion parameters distributed over four different models.
Thanks to this, this platform would potentially be between 10 and 15 times more powerful and precise than ChatGPT.